view tests/test-add.t @ 46607:e9901d01d135

revlog: add a mechanism to verify expected file position before appending

If someone uses `hg debuglocks`, or some non-hg process writes to the .hg
directory without respecting the locks, or if the repo's on a networked
filesystem, it's possible for the revlog code to write out corrupted data.

The form of this corruption can vary depending on what data was written and how
that happened. We are in the "networked filesystem" case (though I've had users
also do this to themselves with the "`hg debuglocks`" scenario), and most often
see this with the changelog. What ends up happening is we produce two items
(let's call them rev1 and rev2) in the .i file that have the same linkrev,
baserev, and offset into the .d file, while the data in the .d file is appended
properly. rev2's compressed_size is accurate for rev2, but when we go to
decompress the data in the .d file, we use the offset that's recorded in the
index file, which is the same as rev1, and attempt to decompress
rev2.compressed_size bytes of rev1's data. This usually does not succeed. :)

When using inline data, this also fails, though I haven't investigated why too
closely. This shows up as a "patch decode" error. I believe what's happening
there is that we're basically ignoring the offset field, getting the data
properly, but since baserev != rev, it thinks this is a delta based on rev
(instead of a full text) and can't actually apply it as such.

For now, I'm going to make this an optional component and default it to entirely
off. I may increase the default severity of this in the future, once I've
enabled it for my users and we gain more experience with it. Luckily, most of my
users have a versioned filesystem and can roll back to before the corruption has
been written, it's just a hassle to do so and not everyone knows how (so it's a
support burden). Users on other filesystems will not have that luxury, and this
can cause them to have a corrupted repository that they are unlikely to know how
to resolve, and they'll see this as a data-loss event. Refusing to create the
corruption is a much better user experience.

This mechanism is not perfect. There may be false-negatives (racy writes that
are not detected). There should not be any false-positives (non-racy writes that
are detected as such). This is not a mechanism that makes putting a repo on a
networked filesystem "safe" or "supported", just *less* likely to cause
corruption.

Differential Revision: https://phab.mercurial-scm.org/D9952

  $ hg init a
  $ cd a
  $ echo a > a
  $ hg add -n
  adding a
  $ hg st
  ? a
  $ hg add
  adding a
  $ hg st
  A a
  $ hg forget a
  $ hg add
  adding a
  $ hg forget a
  $ hg add --color debug
  [ui.addremove.added ui.status|adding a]
  $ hg st
  A a
  $ mkdir dir
  $ cd dir
  $ hg add ../a
  ../a already tracked!
  $ cd ..

  $ echo b > b
  $ hg add -n b
  $ hg st
  A a
  ? b
  $ hg add b
  $ hg st
  A a
  A b

should fail

  $ hg add b
  b already tracked!
  $ hg st
  A a
  A b

#if no-windows
  $ echo foo > con.xml
  $ hg --config ui.portablefilenames=jump add con.xml
  config error: ui.portablefilenames value is invalid ('jump')
  [30]
  $ hg --config ui.portablefilenames=abort add con.xml
  abort: filename contains 'con', which is reserved on Windows: con.xml
  [10]
  $ hg st
  A a
  A b
  ? con.xml
  $ hg add con.xml
  warning: filename contains 'con', which is reserved on Windows: con.xml
  $ hg st
  A a
  A b
  A con.xml
  $ hg forget con.xml
  $ rm con.xml
#endif

#if eol-in-paths
  $ echo bla > 'hello:world'
  $ hg --config ui.portablefilenames=abort add
  adding hello:world
  abort: filename contains ':', which is reserved on Windows: 'hello:world'
  [10]
  $ hg st
  A a
  A b
  ? hello:world
  $ hg --config ui.portablefilenames=ignore add
  adding hello:world
  $ hg st
  A a
  A b
  A hello:world
#endif

  $ hg ci -m 0 --traceback

  $ hg log -r "heads(. or wdir() & file('**'))"
  changeset:   0:* (glob)
  tag:         tip
  user:        test
  date:        Thu Jan 01 00:00:00 1970 +0000
  summary:     0
  
should fail

  $ hg add a
  a already tracked!

  $ echo aa > a
  $ hg ci -m 1
  $ hg up 0
  1 files updated, 0 files merged, 0 files removed, 0 files unresolved
  $ echo aaa > a
  $ hg ci -m 2
  created new head

  $ hg merge
  merging a
  warning: conflicts while merging a! (edit, then use 'hg resolve --mark')
  0 files updated, 0 files merged, 0 files removed, 1 files unresolved
  use 'hg resolve' to retry unresolved file merges or 'hg merge --abort' to abandon
  [1]
  $ hg st
  M a
  ? a.orig

wdir doesn't cause a crash, and can be dynamically selected if dirty

  $ hg log -r "heads(. or wdir() & file('**'))"
  changeset:   2147483647:ffffffffffff
  parent:      2:* (glob)
  parent:      1:* (glob)
  user:        test
  date:        * (glob)
  
should fail

  $ hg add a
  a already tracked!
  $ hg st
  M a
  ? a.orig
  $ hg resolve -m a
  (no more unresolved files)
  $ hg ci -m merge

Issue683: peculiarity with hg revert of an removed then added file

  $ hg forget a
  $ hg add a
  $ hg st
  ? a.orig
  $ hg rm a
  $ hg st
  R a
  ? a.orig
  $ echo a > a
  $ hg add a
  $ hg st
  M a
  ? a.orig

excluded file shouldn't be added even if it is explicitly specified

  $ hg add a.orig -X '*.orig'
  $ hg st
  M a
  ? a.orig

Forgotten file can be added back (as either clean or modified)

  $ hg forget b
  $ hg add b
  $ hg st -A b
  C b
  $ hg forget b
  $ echo modified > b
  $ hg add b
  $ hg st -A b
  M b
  $ hg revert -qC b

  $ hg add c && echo "unexpected addition of missing file"
  c: * (glob)
  [1]
  $ echo c > c
  $ hg add d c && echo "unexpected addition of missing file"
  d: * (glob)
  [1]
  $ hg st
  M a
  A c
  ? a.orig
  $ hg up -C
  1 files updated, 0 files merged, 0 files removed, 0 files unresolved

forget and get should have the right order: added but missing dir should be
forgotten before file with same name is added

  $ echo file d > d
  $ hg add d
  $ hg ci -md
  $ hg rm d
  $ mkdir d
  $ echo a > d/a
  $ hg add d/a
  $ rm -r d
  $ hg up -C
  1 files updated, 0 files merged, 0 files removed, 0 files unresolved
  $ cat d
  file d

Test that adding a directory doesn't require case matching (issue4578)
#if icasefs
  $ mkdir -p CapsDir1/CapsDir
  $ echo abc > CapsDir1/CapsDir/AbC.txt
  $ mkdir CapsDir1/CapsDir/SubDir
  $ echo def > CapsDir1/CapsDir/SubDir/Def.txt

  $ hg add capsdir1/capsdir
  adding CapsDir1/CapsDir/AbC.txt
  adding CapsDir1/CapsDir/SubDir/Def.txt

  $ hg forget capsdir1/capsdir/abc.txt

  $ hg forget capsdir1/capsdir
  removing CapsDir1/CapsDir/SubDir/Def.txt

  $ hg add capsdir1
  adding CapsDir1/CapsDir/AbC.txt
  adding CapsDir1/CapsDir/SubDir/Def.txt

  $ hg ci -m "AbCDef" capsdir1/capsdir

  $ hg status -A capsdir1/capsdir
  C CapsDir1/CapsDir/AbC.txt
  C CapsDir1/CapsDir/SubDir/Def.txt

  $ hg files capsdir1/capsdir
  CapsDir1/CapsDir/AbC.txt
  CapsDir1/CapsDir/SubDir/Def.txt

  $ echo xyz > CapsDir1/CapsDir/SubDir/Def.txt
  $ hg ci -m xyz capsdir1/capsdir/subdir/def.txt

  $ hg revert -r '.^' capsdir1/capsdir
  reverting CapsDir1/CapsDir/SubDir/Def.txt

The conditional tests above mean the hash on the diff line differs on Windows
and OS X
  $ hg diff capsdir1/capsdir
  diff -r * CapsDir1/CapsDir/SubDir/Def.txt (glob)
  --- a/CapsDir1/CapsDir/SubDir/Def.txt	Thu Jan 01 00:00:00 1970 +0000
  +++ b/CapsDir1/CapsDir/SubDir/Def.txt	* (glob)
  @@ -1,1 +1,1 @@
  -xyz
  +def

  $ hg mv CapsDir1/CapsDir/abc.txt CapsDir1/CapsDir/ABC.txt
  $ hg ci -m "case changing rename" CapsDir1/CapsDir/AbC.txt CapsDir1/CapsDir/ABC.txt

  $ hg status -A capsdir1/capsdir
  M CapsDir1/CapsDir/SubDir/Def.txt
  C CapsDir1/CapsDir/ABC.txt

  $ hg remove -f 'glob:**.txt' -X capsdir1/capsdir
  $ hg remove -f 'glob:**.txt' -I capsdir1/capsdir
  removing CapsDir1/CapsDir/ABC.txt
  removing CapsDir1/CapsDir/SubDir/Def.txt
#endif

  $ cd ..

test --dry-run mode in forget

  $ hg init testdir_forget
  $ cd testdir_forget
  $ echo foo > foo
  $ hg add foo
  $ hg commit -m "foo"
  $ hg forget foo --dry-run -v
  removing foo
  $ hg diff
  $ hg forget not_exist -n
  not_exist: $ENOENT$
  [1]

  $ cd ..

test --interactive mode in forget

  $ hg init interactiveforget
  $ cd interactiveforget
  $ echo foo > foo
  $ hg commit -qAm "foo"
  $ echo bar > bar
  $ hg commit -qAm "bar"
  $ hg forget foo --dry-run -i
  abort: cannot specify both --dry-run and --interactive
  [10]

  $ hg forget foo --config ui.interactive=True -i << EOF
  > ?
  > n
  > EOF
  forget foo [Ynsa?] ?
  y - yes, forget this file
  n - no, skip this file
  s - skip remaining files
  a - include all remaining files
  ? - ? (display help)
  forget foo [Ynsa?] n

  $ hg forget foo bar --config ui.interactive=True -i << EOF
  > y
  > n
  > EOF
  forget bar [Ynsa?] y
  forget foo [Ynsa?] n
  removing bar
  $ hg status
  R bar
  $ hg up -qC .

  $ hg forget foo bar --config ui.interactive=True -i << EOF
  > s
  > EOF
  forget bar [Ynsa?] s
  $ hg st
  $ hg up -qC .

  $ hg forget foo bar --config ui.interactive=True -i << EOF
  > a
  > EOF
  forget bar [Ynsa?] a
  removing bar
  removing foo
  $ hg status
  R bar
  R foo
  $ hg up -qC .

  $ cd ..